Girish P. Saraph April 25, 2005 ELITEX 2005 Dr. Girish P. Saraph Associate Professor Department of Electrical Engineering Indian Institute of Technology Bombay Powai, Mumbai Converged Communication Networks

Girish P. Saraph April 25, 2005 ELITEX 2005 Outline Convergence in core networks New broadband applications and requirements Rationale for MPLS traffic engineering New VS routing for MPLS path selection Technical merits of Virtual Space routing MPLS traffic engineering implementation Conclusions

Girish P. Saraph April 25, 2005 ELITEX 2005 Traditional Technologies in Core Networks ATM or Frame Relay network: Optimized for voice transport Advantages: connection oriented, reliable, supports QoS Disadvantages: limited scalability and flexibility, high overhead IP network: Optimized for data (packet) transport Advantages: excellent scalability and flexibility, efficient, common application platform, supports several data services Disadvantages: connectionless, best-effort, no performance guarantees Both networks are supported over TDM or SONET/SDH platform SONET/SDH network: Statically provisioned, reliable (SLA support)

Girish P. Saraph April 25, 2005 ELITEX 2005 MPLS for Convergence in Core Networks Very costly to keep separate voice, video, and data networks Bridging the gap between IP (data) and ATM (voice) networks  MPLS network to support voice, video, and data services Individual label-switched paths with QoS guarantees for aggregate flows – Explicit routed LSPs with specific resource reservation Evolutionary path to IP and ATM infrastructure  Reduced CAPEX Single operation and management plane  Reduced OPEX Additional services over existing networks  Increased ROI – Enable layer 2 and layer 3 virtual private networks (VPNs)

Girish P. Saraph April 25, 2005 ELITEX 2005 Emerging Broadband Applications Web-hosting and Data-warehousing services Internet gaming and streaming video services Video-on-demand and Bandwidth-on-demand services Bundled VoIP, Video telephony, and Wireless access services Video conferencing services (MPLS VPN Multicast) for global enterprises Banking, ERP, and CRM applications of global enterprises Bundled services for digital home network access, including broadband Internet, voice & video telephony, digital TV or HDTV  Different requirements for real-time performance, reliability, security, etc.

Girish P. Saraph April 25, 2005 ELITEX 2005 Broadband Applications Statistical Multiplexing ~ 6250 audio (G.729) channels with ~ 9.6 kbps rate on OC-3 (155Mbps) Link ~ 33 MPEG video channels with ~ 1.82 Mbps rate Analytical Model Broadband applications  Reduced smoothening  Bursty, unpredictable aggregate traffic

Girish P. Saraph April 25, 2005 ELITEX 2005 Rationale for MPLS Traffic Engineering Broadband applications  bursty, unpredictable aggregate traffic  high peak-to-average ratios in link loading Common approach – Static over-provisioning with load-balancing e.g. 2 links carrying < 40% average traffic – Not an acceptable QoS solution under a link failure Achieve high network utilization with overlapping path protection Support QoS from the user perspective using application based aggregation (FEC) and traffic engineering Intelligent path selection based on specific QoS requirements, network resources, and performance parameters

Girish P. Saraph April 25, 2005 ELITEX 2005 Multi-parameter Optimization Traffic engineering objectives require multi-parameter optimization based on: Network performance parameters: Total network throughput (total connections x data rates) Number of hops in each path (leads to network loading) Link loading distribution (congestion and under-utilization of links) Blocking probability for arbitrary connection-requests Stability and scalability of implementation Flow or connection based QoS parameters: Available bandwidthPriority level (or CoS) Total path delayDelay jitter Packet drop ratePath attributes Path protectionReliability

Girish P. Saraph April 25, 2005 ELITEX 2005 New Approach: VS Routing Minimize routing information – concise representation – elimination of redundant info.  better scalability, fast adaptability, and simple implementation Transform network topology information (VS embedding) into multi-dimensional network map (VS configuration) capable of geometric routing (directivity property) Combine directivity and dynamic link info. for path selection (VS routing) Use VS routing for multiple path selection & multi-parameter optimization

Girish P. Saraph April 25, 2005 ELITEX 2005 Virtual Space Routing Differentiate between static and dynamic information Static → Network topology, Dynamic → Link status & loading Static (topology) information is embedded into VS configuration VS configuration enables simple geometric routing Directed distance to destination gives available path choices VS Embedding: Evolution of multi-dimensional VS configuration Energy minimization process in equivalent multi-body system Dynamic link/node costs: Reflects link/node status & loading Link or node failure represents extreme link or node cost Link / node costs are only distributed locally in VS space

Girish P. Saraph April 25, 2005 ELITEX 2005 Technical Merits => VS routing scheme is highly scalable, dynamic, robust, and simple. Traditional scheme VS scheme Routing information scaling: N 2 – N 3 log(N) Information database: Thousands of strings < 100 numbers Peak link loading: – ~ 25% less Link state update time: Slow Fast Multiple QoS parameter support: Partial Full Resource requirements: High Low Failure recovery & convergence: Slow Fast

Girish P. Saraph April 25, 2005 ELITEX 2005 MPLS Traffic Engineering Tool Implementation MPLS-TE Database VS Routing based Path Selection User Interface or Automated Network Management Explicit Route Object Module RSVP Daemon Admin / Policy Control Module Network Diagnostics Network Packet Classifier Using Virtual Space (VS) routing for MPLS-TE optimization

Girish P. Saraph April 25, 2005 ELITEX 2005 MPLS Traffic Engineering & Network Management QOS support for VVD convergence in the core Capability to handle hundreds of MPLS tunnels Path selection based on multiple QoS parameters: (i) hops (ii) delay (iii) bandwidth (iv) congestion (v) link status Efficient overlapping back-up paths for MPLS Fast Reroute Optimal point-to-multipoint LSP tree selection for MPLS multicast Identification of critical links and analysis of “what-if” scenarios Enhanced network resource utilization and performance Improved network management and planning

Girish P. Saraph April 25, 2005 ELITEX 2005 Conclusions Convergence of VVD services in the core is achieved using MPLS network MPLS traffic engineering enables QoS support and network optimization MPLS-TE tool is being developed based on new VS routing scheme Tool supports optimization of multiple QoS parameters, protection, and multicast Acknowledgement: D.I.T. has funded a project at IIT Bombay to carry out performance analysis of new VS routing on the Network Simulator platform. The DIT coordinator is Shri. B.M. Baveja. Ref.: [1] High Performance Switching & Routing (HPSR) 2003, Torino, Italy [2] International Conf. on Communications (ICC) 2004, Paris, France [3] MPLS World Congress 2005, Paris, France